General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

Users may download and print one copy of any publication from the public portal for the purpose of private study or research.

You may not further distribute the material or use it for any profit-making activity or commercial gain

You may freely distribute the URL identifying the publication in the public portal If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from orbit.dtu.dk on: Nov 13, 2021

Separation of Peptides with Forward Osmosis Biomimetic Membranes

Bajraktari, Niada; Madsen, Henrik T; Gruber, Mathias Felix; Truelsen, Sigurd Friis; Jensen, Elzbieta L;Jensen, Henrik; Hélix-Nielsen, Claus

Published in:Membranes

Link to article, DOI:10.3390/membranes6040046

Publication date:2016

Document VersionPublisher's PDF, also known as Version of record

Link back to DTU Orbit

Citation (APA):Bajraktari, N., Madsen, H. T., Gruber, M. F., Truelsen, S. F., Jensen, E. L., Jensen, H., & Hélix-Nielsen, C.(2016). Separation of Peptides with Forward Osmosis Biomimetic Membranes. Membranes, 6(4), [46].https://doi.org/10.3390/membranes6040046

membranes

Article

Separation of Peptides with Forward OsmosisBiomimetic MembranesNiada Bajraktari 1,*, Henrik T. Madsen 2, Mathias F. Gruber 1, Sigurd Truelsen 1,Elzbieta L. Jensen 3, Henrik Jensen 3 and Claus Hélix-Nielsen 1,4,*

1 Department of Environmental Engineering, Technical university of Denmark,Kongens Lyngby 2800, Denmark; nano.mathias@gmail.com (M.F.G.); sigut@env.dtu.dk (S.T.)

2 Department of Chemistry and Bioscience, Aalborg University, Copenhagen 2450, Denmark; htm@bio.aau.dk3 Department of Pharmacy, University of Copenhagen, Copenhagen 2100, Denmark;

ellu57@gmail.com (E.L.J.); henrik.jensen@sund.ku.dk (H.J.)4 Department of Chemical Technology, University of Maribor, Maribor 2000, Slovenia* Correspondence: nbaj@env.dtu.dk (N.B.); clhe@env.dtu.dk (C.H.-N.);

Tel.: +45-45251506 (N.B.); +45-27102076 (C.H.-N.)

Academic Editor: Chuyang Y. TangReceived: 9 September 2016; Accepted: 10 November 2016; Published: 15 November 2016

Abstract: Forward osmosis (FO) membranes have gained interest in several disciplines for therejection and concentration of various molecules. One application area for FO membranes that isbecoming increasingly popular is the use of the membranes to concentrate or dilute high valuecompound solutions such as pharmaceuticals. It is crucial in such settings to control the transportover the membrane to avoid losses of valuable compounds, but little is known about the rejectionand transport mechanisms of larger biomolecules with often flexible conformations. In this study,transport of two chemically similar peptides with molecular weight (Mw) of 375 and 692 Da across athin film composite Aquaporin Inside™ Membrane (AIM) FO membrane was investigated. Despitethe relative large size, both peptides were able to permeate the dense active layer of the AIMmembrane and the transport mechanism was determined to be diffusion-based. Interestingly, themembrane permeability increased 3.65 times for the 692 Da peptide (1.39 × 10−12 m2·s−1) comparedto the 375 Da peptide (0.38 × 10−12 m2·s−1). This increase thus occurs for an 85% increase in Mw butonly for a 34% increase in peptide radius of gyration (Rg) as determined from molecular dynamics(MD) simulations. This suggests that Rg is a strong influencing factor for membrane permeability.Thus, an increased Rg reflects the larger peptide chains ability to sample a larger conformationalspace when interacting with the nanostructured active layer increasing the likelihood for permeation.

Keywords: forward osmosis; biomimetic; peptides; rejection

1. Introduction

Within the membrane community there has been an increasing focus on the filtration techniqueforward osmosis (FO), where filtration is driven by a concentration gradient as opposed to thetraditional pressure gradient known from reverse osmosis (RO). FO membranes have gainedinterest in several disciplines and are finding applications in different markets such as seawater andbrackish water desalination [1,2], wastewater treatment [3–5], treatment of high salinity waters [6–8],fertigation [9,10], textile industry [11,12], dairy [13,14], food [15], and beverage [16,17], powergeneration [18,19], and pharma industry [20,21].

One of the key performance indicators for FO membranes is the rejection of organic molecules.In drinking water or wastewater treatment, high rejection is important to ensure that micropollutantssuch as pesticides, pharmaceuticals, and endocrine disruptors do not end up in the final product.In concentration or dilution applications, such as downstream processing of pharmaceuticals and

Membranes 2016, 6, 46; doi:10.3390/membranes6040046 www.mdpi.com/journal/membranes

Membranes 2016, 6, 46 2 of 12

fertigation, it is also important to control the transport of other molecules than water across themembrane to avoid loss of valuable components.

Previous studies have focused on the rejection of relatively small organic molecules (MW < 365 Da)and with a relatively well defined and rigid structure [22]. Therefore, very little knowledge is availablefor the rejection of larger, more complex molecules. Based on the reported results in previous studies,very high or complete rejection should be expected when filtrating larger molecules. Typically, rejectionhas been found to be size dependent with higher rejections for larger molecules [22,23]. Size dependentrejection indicates that steric hindrance is the underlying rejection mechanism during filtration andtherefore complete rejection should be possible to obtain. In reality however, rejections are notfound to be complete but instead only approximately complete with rejection values in the range of99%–99.9%. This is due to the diffusion of molecules through the membrane matrix, where moleculesfirst adsorb to the active layer, then diffuse across it until they desorb on the support side of theactive layer. Recent findings report that for some membranes, especially dense thin film composite(TFC) membranes, rejection is solely due to diffusion and not steric hindrance [23]. However, at themolecular level, diffusion through the membrane matrix must also be a steric process. If moleculescan diffuse through the membrane matrix, they must be able to move in between the polymer chainsthat constitute the selective layer. Following this logic, there could be an upper limit to the size ofmolecules that are able to diffuse through the membrane and therefore it is interesting to investigatewhether larger molecules with complex structures, such as peptides or proteins, are still able to diffusethrough the membrane matrix or if they become entirely retained. Ultimately, this will have a largeimpact not only on the chosen operating conditions of the system but it will also question the stabilityand robustness of the fabrication process of biomimetic membranes with essential components in theactive layer such as proteins, peptides, polymers, and vesicles.

In this study, the rejection and transport mechanism through the membrane barrier oftwo chemically similar peptides of different molecular size (375 and 692 Da) were investigated with abiomimetic, high rejection, and dense TFC membrane. In an earlier study, the rejection mechanism forthe membrane had been found to be diffusion based, even for small molecules (MW = 149 Da) and themembrane is as such ideal for the purpose of this study [23]. To investigate whether the peptides couldmove through the membrane and if the transport was due to diffusion, the membranes were exposedto different concentration gradients of peptides in a specially designed FO cell. Furthermore, beingcomplex molecules with no rigid structures, a radius of gyration for each peptide was obtained throughmolecular dynamic simulations, which then that radius was related to the peptides permeabilitythrough the membrane matrix.

2. Theory

2.1. FO Filtration

In FO, water transport is driven by an osmotic pressure gradient, ∆π [bar], between the feedand draw solution. For an ideal membrane, the water flux, Jw [L·m−2·h−1], can be described byEquation (1).

Jw = Lp∆π (1)

where Lp [L·m−2·h−1·bar−1] is the membrane permeability coefficient of the membrane.For the same membrane, salt will diffuse from the draw to the feed solution as a result of the salt

concentration gradient, ∆C [g·L−1], as described by Equation (2).

Js = −B∆C (2)

where B [L·m−2·h−1] is the salt permeability coefficient and Js [g·m−2·h−1] is the salt flux.

Membranes 2016, 6, 46 3 of 12

2.2. Separation Performance

The rejection of the solutes in an FO process is described by Equation (3).

R =

(1 −

cp

c f

)100% (3)

where R [%] is the rejection of the specific solute, cf is the average concentration of the solute in thefeed during experiments and cp is the solute concentration in the permeate. The average concentrationof the solute in the feed during experiments is described by Equation (4).

c f =c0 + cend

2(4)

where c0 is the initial concentration of the solute at the start of the experiment and cend is theconcentration at the end of the experiment. Any concentration unit can be used.

2.3. Modeling of Membrane Performance

In a previous study with the same membrane, the rejection of three small organic molecules(atrazine, BAM, and DEIA; 215.69 Da, 190.03 Da, and 145.55 Da, respectively) was modeled with asolution diffusion based model, while a pore flow model could not be fitted [23]. Since the peptides arelarger molecules than the previously investigated pesticides and the hydrodynamic model being basedon size exclusion, it was assumed that that a pore-flow model would not be applicable for the peptides.Instead, the rejection mechanism was investigated with the Fickian’ Solution-Diffusion model [24,25].

In the solution-diffusion model, the membrane is viewed as a solid barrier, through which themolecules move by first adsorbing to the surface and then diffusing through the membrane matrixuntil they reach the other side of the membrane from where they desorb and move into the permeate.The flux through the membrane is described by Equation (5).

Jp =Pl

∆Cp (5)

where Jp [mol·m−2·s−1] is the flux of peptides, l [m] is the membrane thickness, P [m2·s−1] is thepermeability and ∆Cp [mol·m−3] is the average concentration difference across the membrane.

The permeability coefficient can further be described as a function of the diffusion coefficientof the peptide through the membrane and the sorption coefficient of the peptide to the membranematerial as shown in Equation (6).

P = DS (6)

where D [m2·s−1] is the diffusion coefficient for the peptide in the membrane matrix and S is thesorption coefficient of the peptide to the membrane material. S is dimensionless, since it is the ratioof the peptide activity in solution relative to the membrane surface. In practice, it is sufficient todetermine the permeability coefficient when modeling the rejection.

Experimentally, the permeability coefficient for each peptide was determined by measuring theflux of peptides across the membrane at different concentration difference values in the absence ofwater flux. In the assays, the water flux was eliminated by removing the osmotic gradient over themembrane, which in practice was done by using a solution of the same osmolarity on both sides ofthe membrane. The peptide flux was then determined by measuring the peptide concentration in thepermeate after a given amount of time, and the permeability coefficient could be found as the slope ofthe linear plot.

Membranes 2016, 6, 46 4 of 12

3. Experimental

3.1. Forward Osmosis Membranes

Biomimetic FO membranes from Aquaporin A/S were used to study the diffusion mechanismsthrough the membrane. The membranes contain functional aquaporin channels embedded in vesiclesof 200 nm size which are further incorporated in the membrane through a thin film composite activelayer formation [26]. The TFC layer is synthetized on a porous support for mechanical strength duringthe operating procedure.

3.2. Selected Peptides

Two different peptides were used for the up-concentration assay and transport mechanismsexperiments, one larger peptide (GGG SGA GKT) of 692 Da and a second smaller peptide (AGKT) of375 Da, see Figure 1. The two peptides have been tested with both experimental and computationalmethods. The peptides are expected to be very flexible due to the lack of a rigid secondary structure,making their geometrical size difficult to determine. Thus, the radius of gyration is used as a statisticalmeasure for the size of the peptides, a method that is used to represent the size of macromolecules [27].Using Molecular Dynamics (MD), see Computational methods section, the radius of gyration iscalculated based on the most often occurring conformations. The radius of gyration is then correlatedto the permeability of the peptide through the membrane matrix obtained in experimental assays.

Membranes 2016, 6, 46 4 of 12

3. Experimental

3.1. Forward Osmosis Membranes

Biomimetic FO membranes from Aquaporin A/S were used to study the diffusion mechanisms

through the membrane. The membranes contain functional aquaporin channels embedded in vesicles

of 200 nm size which are further incorporated in the membrane through a thin film composite active

layer formation [26]. The TFC layer is synthetized on a porous support for mechanical strength during

the operating procedure.

3.2. Selected Peptides

Two different peptides were used for the up-concentration assay and transport mechanisms

experiments, one larger peptide (GGG SGA GKT) of 692 Da and a second smaller peptide (AGKT) of

375 Da, see Figure 1. The two peptides have been tested with both experimental and computational

methods. The peptides are expected to be very flexible due to the lack of a rigid secondary structure,

making their geometrical size difficult to determine. Thus, the radius of gyration is used as a statistical

measure for the size of the peptides, a method that is used to represent the size of macromolecules

[27]. Using Molecular Dynamics (MD), see Computational methods section, the radius of gyration is

calculated based on the most often occurring conformations. The radius of gyration is then correlated

to the permeability of the peptide through the membrane matrix obtained in experimental assays.

From a practical point of view, the two peptides are interesting since they potentially can be

used to capture and recover phosphate. The peptide sequence Ser-Gly-Ala-Gly-Lys-Thr (SGA GKT)

is a reconstructed P-loop in which proteins are able to bind phosphate groups of GTP and ATP. This

hexapeptide has previously been reported to bind to HPO42− at high pH [28] and may therefore be

able to bind phosphorous compounds as a means to address the future issues with diminishing

phosphorous resources.

Figure 1. Chemical representation of the two peptides. (a) AGKT (375 Da) and (b) GGG SGA GKT

(692 Da). The peptides are constructs based on the amino acid configuration described by Bianchi

et al. [28]. The peptide amine (N–H) backbone groups form a ‘nest’ structure toward phosphate anions

and the Lysine side chain participates in the binding when phosphorous compounds are present.

Color code: green = C, blue = N, red = O, and white = H.

3.3. Forward Osmosis Operating System

An FO operating module was designed, built, and used to conduct the experiments, as shown

in Figure 2. The device consists of two compartments, a static upper unit containing the buffer

solution and a lower unit where the peptide sample was recirculated. The device can also be used for

FO experiments, with one of the units containing a draw solution with osmotic potential for up-

concentration or dilution of samples. In this study, the device was used to conduct diffusion

experiments without the effect of the osmotic gradient, therefore the draw solution was replaced with

a buffer solution. Thus, there was only a peptide concentration gradient over the membrane barrier.

Figure 1. Chemical representation of the two peptides. (a) AGKT (375 Da) and (b) GGG SGAGKT (692 Da). The peptides are constructs based on the amino acid configuration described byBianchi et al. [28]. The peptide amine (N–H) backbone groups form a ‘nest’ structure toward phosphateanions and the Lysine side chain participates in the binding when phosphorous compounds are present.Color code: green = C, blue = N, red = O, and white = H.

From a practical point of view, the two peptides are interesting since they potentially can beused to capture and recover phosphate. The peptide sequence Ser-Gly-Ala-Gly-Lys-Thr (SGA GKT)is a reconstructed P-loop in which proteins are able to bind phosphate groups of GTP and ATP.This hexapeptide has previously been reported to bind to HPO2−

4 at high pH [28] and may thereforebe able to bind phosphorous compounds as a means to address the future issues with diminishingphosphorous resources.

3.3. Forward Osmosis Operating System

An FO operating module was designed, built, and used to conduct the experiments, as shownin Figure 2. The device consists of two compartments, a static upper unit containing the buffersolution and a lower unit where the peptide sample was recirculated. The device can also be usedfor FO experiments, with one of the units containing a draw solution with osmotic potential forup-concentration or dilution of samples. In this study, the device was used to conduct diffusion

Membranes 2016, 6, 46 5 of 12

experiments without the effect of the osmotic gradient, therefore the draw solution was replaced witha buffer solution. Thus, there was only a peptide concentration gradient over the membrane barrier.Membranes 2016, 6, 46 5 of 12

Figure 2. Forward osmosis operating module with static top compartment. (a) Showing the drawings

of the lower unit where the peptide sample was recirculated and (b) showing the assembled module.

The two units are separated by a flat sheet membrane tightened with an o-ring and fastened together

mechanically through screws.

3.4. Rejection Model Assay

The FO module was used to determine the diffusion of the peptide molecules through the

membrane matrix. A vial of each peptide (10 mg) was solubilized in TES buffer (2-[2-Hydroxy-1,1-

bis(hydroxymethyl)ethyl)amino]ethanesulfonic acid) consisting of 10 mM TES, 1 mM EDTA and the

pH was adjusted to 8 with 1 M NaOH to generate stock solutions of 100 mg/L. From this, peptide

solutions of 1 mg/L, 5 mg/L, and 10 mg/L were prepared. 0.5 mL of TES buffer (without peptide) was

placed in the reservoir compartment and 100 mL of the buffered peptide solution was re-circulated

in the bottom compartment. The system was saturated for five hours or overnight to avoid any start

up effects such as the loss of peptide to the chamber walls and the membrane. Before starting the

diffusion experiment, the reservoir compartment containing the pure TES buffer was emptied and

washed with TES buffer. The experiment was initiated by placing a fresh 0.5 mL aliquot of TES buffer

in the reservoir compartment and was then set to run for five hours. At the end of the experiment,

the TES buffer solution was removed from the reservoir compartment, weighed, and analyzed with

HPLC to quantify potential diffusing peptides. The previously described steps were followed for

each of the chosen concentrations for both the peptides.

3.5. Analytical Methods

The peptides were analyzed with HPLC/UV (1200 Infinity, Agilent Technology, Santa Clara, CA,

USA), equipped with a ZORBAX Eclipse XDB-C18 5 µm column (Agilent Technology, Santa Clara,

CA, USA). For GGG SGA GKT, an eluent mixture of acetonitrile and MilliQ (Merck Millipore,

Darmstadt, Germany) in the ratio 10/90 was used. The injection volume was 50 µL and the flow rate

1 mL/min. For AGKT, the peptides were eluted with a 67 mM phosphate buffer at pH 7.4. The

injection volume was 50 µL and the flow rate 0.5 mL/min. The chosen absorption wavelength was

205 nm for both peptides.

3.6. Computational Methods

The chosen peptides have a flexible structure making it difficult to determine their geometrical

size from a single conformational minima. Thus, the radius of gyration is used as a statistical measure

of the geometrical size of the peptides. Molecular Dynamics (MD) was used to determine the radius

of gyration of the peptides based on hierarchical clusters from the generated trajectory. Simulations

were performed with the molecular modeling packages Amber14 and AmberTools14 [29]. Fully

extended peptides were created with LEaP in AmberTools14 using the ff14SB force field parameters

[30] and were explicitly solvated using the TIP3P model for water [31] in rectangular boxes with 10.0

Å to each edge from the peptides. The systems were minimized for 10,000 steps: 5000 steps using the

Figure 2. Forward osmosis operating module with static top compartment. (a) Showing the drawingsof the lower unit where the peptide sample was recirculated and (b) showing the assembled module.The two units are separated by a flat sheet membrane tightened with an o-ring and fastened togethermechanically through screws.

3.4. Rejection Model Assay

The FO module was used to determine the diffusion of the peptide molecules through themembrane matrix. A vial of each peptide (10 mg) was solubilized in TES buffer (2-[2-Hydroxy-1,1-bis(hydroxymethyl)ethyl)amino]ethanesulfonic acid) consisting of 10 mM TES, 1 mM EDTA and thepH was adjusted to 8 with 1 M NaOH to generate stock solutions of 100 mg/L. From this, peptidesolutions of 1 mg/L, 5 mg/L, and 10 mg/L were prepared. 0.5 mL of TES buffer (without peptide) wasplaced in the reservoir compartment and 100 mL of the buffered peptide solution was re-circulatedin the bottom compartment. The system was saturated for five hours or overnight to avoid any startup effects such as the loss of peptide to the chamber walls and the membrane. Before starting thediffusion experiment, the reservoir compartment containing the pure TES buffer was emptied andwashed with TES buffer. The experiment was initiated by placing a fresh 0.5 mL aliquot of TES bufferin the reservoir compartment and was then set to run for five hours. At the end of the experiment, theTES buffer solution was removed from the reservoir compartment, weighed, and analyzed with HPLCto quantify potential diffusing peptides. The previously described steps were followed for each of thechosen concentrations for both the peptides.

3.5. Analytical Methods

The peptides were analyzed with HPLC/UV (1200 Infinity, Agilent Technology, Santa Clara, CA,USA), equipped with a ZORBAX Eclipse XDB-C18 5 µm column (Agilent Technology, Santa Clara, CA,USA). For GGG SGA GKT, an eluent mixture of acetonitrile and MilliQ (Merck Millipore, Darmstadt,Germany) in the ratio 10/90 was used. The injection volume was 50 µL and the flow rate 1 mL/min.For AGKT, the peptides were eluted with a 67 mM phosphate buffer at pH 7.4. The injection volumewas 50 µL and the flow rate 0.5 mL/min. The chosen absorption wavelength was 205 nm forboth peptides.

3.6. Computational Methods

The chosen peptides have a flexible structure making it difficult to determine their geometricalsize from a single conformational minima. Thus, the radius of gyration is used as a statistical measureof the geometrical size of the peptides. Molecular Dynamics (MD) was used to determine the radius ofgyration of the peptides based on hierarchical clusters from the generated trajectory. Simulations were

Membranes 2016, 6, 46 6 of 12

performed with the molecular modeling packages Amber14 and AmberTools14 [29]. Fully extendedpeptides were created with LEaP in AmberTools14 using the ff14SB force field parameters [30] andwere explicitly solvated using the TIP3P model for water [31] in rectangular boxes with 10.0 Å toeach edge from the peptides. The systems were minimized for 10,000 steps: 5000 steps using thesteepest decent algorithm and 5000 steps using a conjugated gradient algorithm. The systems werethen heated to 300 K over 0.8 ns with a weak constraint of 1 kcal/mol on the peptide (NVT ensemble),and subsequently equilibrated for 0.2 ns without any restraints at 300 K (NPT ensemble). Productionsimulations were run for 800 ns using the Langevin thermostat with a collision frequency of 1 ps−1,and the pressure was set to 1 bar using the Berendsen barostat. Non-bonded interactions were cut offat 10.0 Å, full electrostatics for the periodic system were calculated using the Particle Mesh Ewaldapproach [32], and hydrogen bonds were restrained using SHAKE [33].

The mass weighted radius of gyration, Rg [Å] from a molecular configuration can be calculated asshown in Equation (7).

R2g =

1M ∑

imi(Ri − RCM)2 (7)

where M is the total molecular mass [Da], mi [Da] is the mass of the ith atom with position vector Ri[Å]. RCM [Å] is the center of mass and calculated as shown in Equation (8).

RCM =1M ∑

imiRi (8)

To compare the influence of a rigid secondary structure on permeability, the radius of gyrationwas determined for three pesticides from a previous study [23]. The radius of gyration for both thepesticides and peptides is calculated on all heavy atoms by the use of CPPTRAJ [34].

Table 1 provides an overview of the comparison between the peptides and the smaller pesticides.Here, the permeabilites have been used to calculate a rejection value using Equation (9), owing to thefact that rejection values are more commonly used in membrane studies to indicate the retention of agiven compound by the membrane.

R =

(1 − P

t × Jw

)× 100% (9)

where t is the membrane thickness (m) and Jw the water flux (m3·m−2·s−1). The equation assumesthat feed and retentate concentration are equal, appropriate for low recovery values typically usedto determine rejection values and that the concentration difference across the membrane can beapproximated to be equal to the feed concentration, which is a good approximation for high rejectionvalues (R > 95%).

Table 1. Overview and comparison of membrane transport parameters. The table shows characterizationparameters for each peptide and a comparison to three pesticides in a previous study with the samemembrane [23]. The rejection values are calculated on measured permeabilities and a water flux of9.71 L·m−2·h−1 reported in the previous study. The membrane thickness was measured to be 112 µmand the active area for the setup was 0.785 cm2. The flux is based on a feed concentration of 1 mg·L−1.

Compound Permeability(m2·s−1)

Molecular Weight(g·mol−1)

Radius of Gyration(Å)

Flux(µg·m−2·h−1)

Rejection(%)

Peptide

AGKT 0.38 × 10−12 ± 1.14 × 10−12 375 4.4 ± 0.1 12 ± 37 99.9 ± 0.4GGG SGA GKT 1.39 × 10−12 ± 0.9 × 10−12 692 5.9 ± 0.4 45 ± 29 99.5 ± 0.3

Pesticide

DEIA 5.36 × 10−12 146 1.24 172 98.2BAM 3.9 × 10−12 190 1.30 125 98.7

Atrazine 4.31 × 10−12 216 1.73 139 98.6

Membranes 2016, 6, 46 7 of 12

3.7. Data Analysis

To determine the rejection mechanisms, the permeate concentrations from the HPLC measurementswere used to determine the molar flux of the peptides. The concentration of the peptide was multipliedwith the volume in the draw chamber (0.5 mL) and divided with the membrane area (0.785 cm2)to generate the peptide flux in mol·s−1·m−2. The molar flux was then plotted as a function of theconcentration difference across the membrane. A linear plot shows a diffusion driven process wherethe slope is the permeability coefficient, P, from Equation (6).

4. Results and Discussion

4.1. Peptide Transport Mechanism

The peptides were found to be able to pass the membrane, although in small amounts. Roughlyless than 1% of the peptides were observed to pass the membrane barrier. From the plots in Figure 3it can be seen that the molar flux of the peptides is linearly correlated to the concentration gradientacross the membrane. This strongly indicates that the transport mechanism is based on diffusion asdescribed by the solution-diffusion model.

Membranes 2016, 6, 46 7 of 12

membrane area (0.785 cm2) to generate the peptide flux in mol·s−1·m−2. The molar flux was then plotted

as a function of the concentration difference across the membrane. A linear plot shows a diffusion

driven process where the slope is the permeability coefficient, P, from Equation (6).

4. Results and Discussion

4.1. Peptide Transport Mechanism

The peptides were found to be able to pass the membrane, although in small amounts. Roughly

less than 1% of the peptides were observed to pass the membrane barrier. From the plots in Figure 3

it can be seen that the molar flux of the peptides is linearly correlated to the concentration gradient

across the membrane. This strongly indicates that the transport mechanism is based on diffusion as

described by the solution-diffusion model.

Figure 3. Molar flux of peptides across the membrane. The graph shows the molar peptide flux over

the membrane as a function of the concentration gradient across the membrane. Each concentration

point is a mean value of triplicate experiments at that concentration. The slope of the graph gives the

membrane permeability for each peptide. The linear regression coefficients are 0.99792 (375 Da) and

0.99817 (692 Da).

Interestingly, the larger peptide was found to have a higher permeability. Based on steric

considerations, the larger peptide would be expected to show a lower permeability, since its larger

size would make it more difficult to diffuse through the membrane matrix. Furthermore, similar

constructs of the peptides have previously been described to exhibit flexible structures, where the

peptides can assume several different conformations. To further investigate whether the permeability

is affected by the flexibility of the peptides, MD simulations were performed.

4.2. MD Analysis of Peptide Structure

Molecular dynamic simulations were run to determine the radius of gyration of each peptide

respectively. Hierarchical clusters of the peptides are shown in Figure 4.

Figure 3. Molar flux of peptides across the membrane. The graph shows the molar peptide flux overthe membrane as a function of the concentration gradient across the membrane. Each concentrationpoint is a mean value of triplicate experiments at that concentration. The slope of the graph gives themembrane permeability for each peptide. The linear regression coefficients are 0.99792 (375 Da) and0.99817 (692 Da).

Interestingly, the larger peptide was found to have a higher permeability. Based on stericconsiderations, the larger peptide would be expected to show a lower permeability, since its larger sizewould make it more difficult to diffuse through the membrane matrix. Furthermore, similar constructsof the peptides have previously been described to exhibit flexible structures, where the peptides canassume several different conformations. To further investigate whether the permeability is affected bythe flexibility of the peptides, MD simulations were performed.

4.2. MD Analysis of Peptide Structure

Molecular dynamic simulations were run to determine the radius of gyration of each peptiderespectively. Hierarchical clusters of the peptides are shown in Figure 4.

Membranes 2016, 6, 46 8 of 12

From Figure 4, it can be seen that the peptides are shown to exhibit a disordered nature,assuming several different conformations, with the lysine residue in indeterminate orientations.This is in accordance with a previous study where similar constructs (SGA GKT) of the peptides wereinvestigated by MD [35]. Here the peptide of inspiration (SGA GKT) was studied as peptide-only andpeptide-phosphate systems. The peptide conformational ensemble was stabilized by the presence of aphosphate anion, however the isolated peptide was likely to spend most of the time in a disorderedstate [35]. Thus, it is expected that the peptide will exhibit a flexible structure when in contact with themembrane matrix.

Membranes 2016, 6, 46 8 of 12

From Figure 4, it can be seen that the peptides are shown to exhibit a disordered nature,

assuming several different conformations, with the lysine residue in indeterminate orientations. This

is in accordance with a previous study where similar constructs (SGA GKT) of the peptides were

investigated by MD [35]. Here the peptide of inspiration (SGA GKT) was studied as peptide-only and

peptide-phosphate systems. The peptide conformational ensemble was stabilized by the presence of

a phosphate anion, however the isolated peptide was likely to spend most of the time in a disordered

state [35]. Thus, it is expected that the peptide will exhibit a flexible structure when in contact with

the membrane matrix.

Figure 4. Hierarchical clusters of peptide simulations in water. (a) AGKT and (b) GGG SGA GKT. The

backbone is shown in green and the lysine side chain is marked in stick representation. The

illustrations show an ensemble of the same peptides in the many different structural conformations

they can exhibit.

From the molecular dynamic simulations, the radii of gyrations are obtained. These are shown

in Table 1. The GGG SGA GKT peptide has a radius of gyration 34% larger than the AGKT peptide.

The main difference between the two peptides is the additional GGGS amino acids on the larger

peptide, which primarily makes this peptide longer and not wider, since these amino acids have short

side chains. The longer chain provides the GGG SGA GKT peptide with a larger number of degrees

of freedom for conformational operation and for these two peptides the higher radius of gyration of

the larger peptide is as such also correlated to a greater flexibility and ability to bend, which have

been found to play a role in the transport of larger molecules through membranes [36,37].

4.3. Effect of Size, Flexibility, and Polarity in Membrane Transport

Size is still an important descriptor for permeability as shown in Figure 5a. Here the

permeabilities of the two peptides are compared to those determined previously for the smaller

pesticides (DEIA, BAM, and Atrazine) and this plot clearly indicates that there is a general tendency

for larger molecules to show lower permeabilities. The permeability of the pesticides is 5 to 10 times

higher than for the peptides. In Figure 5a, molecular weight is used to represent the size of the

molecules, but molecular weight might not be the best descriptor of size since it fails to take the

complexity of the peptide backbone chains into account. Compared to smaller molecules, it is not

easy to fit large molecules with a simple geometric shape such as a rectangle or a cylinder. To better

accommodate the geometric structure of the peptides, permeabilities have been plotted as a function

of radius of gyration in Figure 5b and it can be seen that this offers a better description. As argued

before, for these peptides, the radius of gyration is correlated to the degree of freedom and this may

be important for large molecules to pass the membrane barrier. The hypothesis is based on the

increase in entropy with an increasing number of conformations, which in practice means that the

larger peptide can twist and bend its way through the membrane layers and ultimately be traced on

the draw side.

Figure 4. Hierarchical clusters of peptide simulations in water. (a) AGKT and (b) GGG SGAGKT. The backbone is shown in green and the lysine side chain is marked in stick representation.The illustrations show an ensemble of the same peptides in the many different structural conformationsthey can exhibit.

From the molecular dynamic simulations, the radii of gyrations are obtained. These are shownin Table 1. The GGG SGA GKT peptide has a radius of gyration 34% larger than the AGKT peptide.The main difference between the two peptides is the additional GGGS amino acids on the largerpeptide, which primarily makes this peptide longer and not wider, since these amino acids have shortside chains. The longer chain provides the GGG SGA GKT peptide with a larger number of degrees offreedom for conformational operation and for these two peptides the higher radius of gyration of thelarger peptide is as such also correlated to a greater flexibility and ability to bend, which have beenfound to play a role in the transport of larger molecules through membranes [36,37].

4.3. Effect of Size, Flexibility, and Polarity in Membrane Transport

Size is still an important descriptor for permeability as shown in Figure 5a. Here the permeabilitiesof the two peptides are compared to those determined previously for the smaller pesticides (DEIA,BAM, and Atrazine) and this plot clearly indicates that there is a general tendency for larger moleculesto show lower permeabilities. The permeability of the pesticides is 5 to 10 times higher than for thepeptides. In Figure 5a, molecular weight is used to represent the size of the molecules, but molecularweight might not be the best descriptor of size since it fails to take the complexity of the peptidebackbone chains into account. Compared to smaller molecules, it is not easy to fit large moleculeswith a simple geometric shape such as a rectangle or a cylinder. To better accommodate the geometricstructure of the peptides, permeabilities have been plotted as a function of radius of gyration inFigure 5b and it can be seen that this offers a better description. As argued before, for these peptides,the radius of gyration is correlated to the degree of freedom and this may be important for largemolecules to pass the membrane barrier. The hypothesis is based on the increase in entropy with anincreasing number of conformations, which in practice means that the larger peptide can twist andbend its way through the membrane layers and ultimately be traced on the draw side.

Membranes 2016, 6, 46 9 of 12Membranes 2016, 6, 46 9 of 12

Figure 5. Permeability as a function of molecular weight (a) and radius of gyration (b). As a

comparison, three previously tested pesticides are shown in ■ symbols. The two peptides are shown

in ● symbols.

Another factor influencing the permeability of the peptide through the membrane is their

polarity and charge. Polarity can influence the sorption of the peptides to the membrane surface,

which would be the first step in a solution-diffusion based transport mechanism, and the orientation

of the molecule towards the membrane surface [38]. Charge may also influence peptide orientation

towards the membrane surface and may lead to either attraction or repulsion depending on the

relative charge between peptide and membrane. Bianchi et al. investigated the charge of the SGA

GKT peptide as a function of pH and, based on these results, half the population of AGKT peptides

are expected to carry a positive charge in these experiments. The same is estimated for the GGG SGA

GKT peptide. The additional amino acids in the peptide structure carry no charge-bearing side

groups and the N-terminal of alanine and glycine have very similar pKa values (9.69 and 9.60) [39].

The charge of the membrane has not been determined in this study. However, the membrane is a

TFC membrane and these membranes typically have a negative surface charge. Based on this and a

previous study where the AQP membrane was found to interact with cationic flocculants, the AQP

membrane is most likely negatively charged. The opposite charge of the membrane surface and the

peptides will allow for attraction. The peptides can as such become ionically bound to the negatively

charged carboxylic acid surface groups and the diffusion through the membrane matrix could then

occur with the peptide jumping from one negatively charged surface group to the next. This might

also be used to explain why the larger peptide has the highest permeability. Compared to the smaller

peptide, it has an additional chain of four amino acids (GGG S), and this may make it more difficult

for the positively charged side group of the larger peptide to get into contact with the negative surface

groups. It would interact less with the membrane surface and its diffusion through the membrane

matrix would be relatively free. In comparison, the AGKT peptide would be slowed down by the

negative surface groups and thus experience a lower permeability.

In Table 1, the 95% confidence interval for the peptide permeability coefficients are also given.

Compared to the permeability coefficients, they are relatively large, and this reflects the heterogeneity

of the membrane surface. The permeability coefficients were determined for several incremental

pieces of the same membrane sheet, and were found to vary significantly between pieces. The

confidence intervals can thus be understood as a measure of the heterogeneousness of the membrane

surface.

The finding, that a molecule like the 692 Da peptide is capable of penetrating a selective layer

that has been found to show almost complete rejection of molecules that are 4.5 times smaller (DEIA,

146 Da) is highly interesting. It shows that for even dense and highly rejecting membranes,

permeation of significantly larger molecules is still possible. This phenomenon in FO filtration might

Figure 5. Permeability as a function of molecular weight (a) and radius of gyration (b). As a comparison,three previously tested pesticides are shown in � symbols. The two peptides are shown in symbols.

Another factor influencing the permeability of the peptide through the membrane is their polarityand charge. Polarity can influence the sorption of the peptides to the membrane surface, which wouldbe the first step in a solution-diffusion based transport mechanism, and the orientation of the moleculetowards the membrane surface [38]. Charge may also influence peptide orientation towards themembrane surface and may lead to either attraction or repulsion depending on the relative chargebetween peptide and membrane. Bianchi et al. investigated the charge of the SGA GKT peptide asa function of pH and, based on these results, half the population of AGKT peptides are expected tocarry a positive charge in these experiments. The same is estimated for the GGG SGA GKT peptide.The additional amino acids in the peptide structure carry no charge-bearing side groups and theN-terminal of alanine and glycine have very similar pKa values (9.69 and 9.60) [39]. The charge ofthe membrane has not been determined in this study. However, the membrane is a TFC membraneand these membranes typically have a negative surface charge. Based on this and a previous studywhere the AQP membrane was found to interact with cationic flocculants, the AQP membrane is mostlikely negatively charged. The opposite charge of the membrane surface and the peptides will allowfor attraction. The peptides can as such become ionically bound to the negatively charged carboxylicacid surface groups and the diffusion through the membrane matrix could then occur with the peptidejumping from one negatively charged surface group to the next. This might also be used to explain whythe larger peptide has the highest permeability. Compared to the smaller peptide, it has an additionalchain of four amino acids (GGG S), and this may make it more difficult for the positively charged sidegroup of the larger peptide to get into contact with the negative surface groups. It would interact lesswith the membrane surface and its diffusion through the membrane matrix would be relatively free.In comparison, the AGKT peptide would be slowed down by the negative surface groups and thusexperience a lower permeability.

In Table 1, the 95% confidence interval for the peptide permeability coefficients are also given.Compared to the permeability coefficients, they are relatively large, and this reflects the heterogeneityof the membrane surface. The permeability coefficients were determined for several incremental piecesof the same membrane sheet, and were found to vary significantly between pieces. The confidenceintervals can thus be understood as a measure of the heterogeneousness of the membrane surface.

The finding, that a molecule like the 692 Da peptide is capable of penetrating a selective layerthat has been found to show almost complete rejection of molecules that are 4.5 times smaller (DEIA,146 Da) is highly interesting. It shows that for even dense and highly rejecting membranes, permeationof significantly larger molecules is still possible. This phenomenon in FO filtration might lead to a slowaccumulation of the molecules in the draw solution, especially if the draw solution is recovered withmembrane distillation and the molecules have a low volatility, which is the standard for biomolecules.The membrane penetration of the compounds might also ultimately alter the membrane performance.

Membranes 2016, 6, 46 10 of 12

This is especially true for compounds that have a higher affinity for the membrane material and assuch move into the structure and accumulate, which may ultimately change membrane characteristics.Also, within the FO community there is a constant search for the holy grail of draw solutes. If potentialcandidates are toxic even in very low concentrations, it is crucial to know that they might cross themembrane barrier to the feed side even when the size of the draw solute is relatively large.

Finally, biomimetic membranes are based on the concept that biologically active moleculesare incorporated in the selective layer of the membrane. This study shows that biomolecules nottoo dissimilar to proteins are capable of diffusing through the membrane matrix. The question isthen at which size does it become completely impossible for these molecules to diffuse through themembrane. Biomimetic membranes and in general mixed matrix composite membranes are based onthe incorporation of particles in the selective layer. The results of this study indicate that ultimatelythe structure of a mixed matrix membrane might not be static, leading to a leakage of membranecomponents to the surrounding medium. One way to overcome this could be to covalently bind theproteins in biomimetic structures to their surrounding environment in the selective layer.

5. Conclusions

In this study, the transport of two peptides of molecular sizes 375 Da and 692 Da withflexible structures across a biomimetic forward osmosis membrane with a dense selective layer wasinvestigated. The membrane exhibited high but incomplete rejection rates where 1% of the peptideswere still able to cross the membrane barrier in spite of their relatively large size. The peptide with thesmallest molecular mass was found to have the lowest permeability, which might be explained by theradius of gyration. A higher radius of gyration enables the peptide to assume several conformationswhen in contact with the membrane matrix, which ultimately could increase the probability of thepeptide being transported through the membrane. The transport mechanism was found to be diffusionbased. The overall finding of this study was that even relatively large molecules can cross an otherwisedense membrane layer. This is significant primarily for choosing the optimal operating conditionsand run time of the system in downstream processing and needs to be considered for each andevery forward osmosis application. Ultimately, it has a significance in the robustness of the membranefabrication process where the incorporation of biomolecules in the selective layer is one of the key steps.

Acknowledgments: This work was supported by the Danish Innovation Foundation via the IBISS grant(097-2012-4).

Author Contributions: Niada Bajraktari planned, conceived, designed, and built the forward osmosis operatingmodule. Additionally, Niada Bajraktari planned, designed, and performed all experiments; analyzed the data;made the graphs and figures; and was the main author of the manuscript. Henrik T. Madsen laid the groundworkin the previous study with pesticides, assisted in data treatment for solution-diffusion, and co-wrote the paper.Mathias F. Gruber performed the MD simulations on peptides and Sigurd Truelsen calculated radii of gyrations forpesticides. Elzbieta L. Jensen and Henrik Jensen performed the HPLC analysis. Claus Hélix-Nielsen supervisedthe work and gave feedback during the writing of the article.

Conflicts of Interest: The authors declare no conflict of interest. The founding sponsors had no role in the designof the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in thedecision to publish the results.

References

1. Zhao, S.; Zou, L.; Mulcahy, D. Brackish water desalination by a hybrid forward osmosis–nanofiltrationsystem using divalent draw solute. Desalination 2012, 284, 175–181. [CrossRef]

2. Zhang, Y.; Pinoy, L.; Meesschaert, B.; van der Bruggen, B. A natural driven membrane process for brackishand wastewater treatment: Photovoltaic powered ED and FO hybrid system. Environ. Sci. Technol. 2013, 47,10548–10555. [CrossRef] [PubMed]

3. Coday, B.D.; Xu, P.; Beaudry, E.G.; Herron, J.; Lampi, K.; Hancock, N.T.; Cath, T.Y. The sweet spot of forwardosmosis: Treatment of produced water, drilling wastewater, and other complex and difficult liquid streams.Desalination 2014, 333, 23–35. [CrossRef]

Membranes 2016, 6, 46 11 of 12

4. Aftab, B.; Khan, S.J.; Maqbool, T.; Hankins, N.P. High strength domestic wastewater treatment withsubmerged forward osmosis membrane bioreactor. Water Sci. Technol. 2015, 72, 141–149. [CrossRef] [PubMed]

5. Hey, T.; Bajraktari, N.; Vogel, J.; Hélix Nielsen, C.; la Cour Jansen, J.; Jönsson, K. The effects of physicochemicalwastewater treatment operations on forward osmosis. Environ. Technol. 2016, 1–13. [CrossRef] [PubMed]

6. McGinnis, R.L.; Hancock, N.T.; Nowosielski-Slepowron, M.S.; McGurgan, G.D. Pilot demonstration ofthe NH3/CO2 forward osmosis desalination process on high salinity brines. Desalination 2013, 312, 67–74.[CrossRef]

7. Chen, G.; Wang, Z.; Nghiem, L.D.; Li, X.-M.; Xie, M.; Zhao, B.; Zhang, M.; Song, J.; He, T. Treatment of shalegas drilling flowback fluids (SGDFs) by forward osmosis: Membrane fouling and mitigation. Desalination2015, 366, 113–120. [CrossRef]

8. Stone, M.L.; Rae, C.; Stewart, F.F.; Wilson, A.D. Switchable polarity solvents as draw solutes for forwardosmosis. Desalination 2013, 312, 124–129. [CrossRef]

9. Raval, H.D.; Koradiya, P. Direct fertigation with brackish water by a forward osmosis system convertingdomestic reverse osmosis module into forward osmosis membrane element. Desalin. Water Treat. 2015, 3994,1–8. [CrossRef]

10. Phuntsho, S.; Lotfi, F.; Hong, S.; Shaffer, D.L.; Elimelech, M.; Shon, H.K. Membrane scaling and flux declineduring fertiliser-drawn forward osmosis desalination of brackish groundwater. Water Res. 2014, 57, 172–182.[CrossRef] [PubMed]

11. Petrinic, I.; Bajraktari, N.; Hélix-Nielsen, C. Advances in Membrane Technologies for Water Treatment; Elsevier:Amsterdam, The Netherlands, 2015.

12. Ammar, A.; Dofan, I.; Jegatheesan, V.; Muthukumaran, S.; Shu, L. Comparison between nanofiltration andforward osmosis in the treatment of dye solutions. Desalin. Water Treat. 2015, 54, 853–861. [CrossRef]

13. Aydiner, C.; Sen, U.; Topcu, S.; Ekinci, D.; Altinay, A.D.; Koseoglu-Imer, D.Y.; Keskinler, B. Techno-economicviability of innovative membrane systems in water and mass recovery from dairy wastewater. J. Memb. Sci.2014, 458, 66–75. [CrossRef]

14. Pal, P.; Nayak, J. Development and analysis of a sustainable technology in manufacturing acetic acid andwhey protein from waste cheese whey. J. Clean. Prod. 2016, 112, 59–70. [CrossRef]

15. Sant’Anna, V.; Marczak, L.D.F.; Tessaro, I.C. Membrane concentration of liquid foods by forward osmosis:Process and quality view. J. Food Eng. 2012, 111, 483–489. [CrossRef]

16. Cath, T.; Childress, A.; Elimelech, M. Forward osmosis: Principles, applications, and recent developments.J. Memb. Sci. 2006, 281, 70–87. [CrossRef]

17. Jiao, B.; Cassano, A.; Drioli, E. Recent advances on membrane processes for the concentration of fruit juices:A review. J. Food Eng. 2004, 63, 303–324. [CrossRef]

18. Madsen, H.T.; Nissen, S.S.; Søgaard, E.G. Theoretical Framework for Energy Analysis of Hypersaline PressureRetarded Osmosis. Chem. Eng. Sci. 2015, 139, 211–220. [CrossRef]

19. Wan, C.F.; Chung, T.-S. Osmotic power generation by pressure retarded osmosis using seawater brine as thedraw solution and wastewater retentate as the feed. J. Memb. Sci. 2015, 479, 148–158. [CrossRef]

20. Jin, X.; Shan, J.; Wang, C.; Wei, J.; Tang, C.Y. Rejection of pharmaceuticals by forward osmosis membranes.J. Hazard. Mater. 2012, 227–228, 55–61. [CrossRef] [PubMed]

21. Kong, F.; Yang, H.; Wu, Y.; Wang, X.; Xie, Y.F. Rejection of pharmaceuticals during forward osmosis andprediction by using the solution-diffusion model. J. Memb. Sci. 2014, 476, 410–420. [CrossRef]

22. Coday, B.D.; Yaffe, B.G.M.; Xu, P.; Cath, T.Y. Rejection of Trace Organic Compounds by Forward OsmosisMembranes: A Literature Review. Environ. Sci. Technol. 2014, 48, 3612–3624. [CrossRef] [PubMed]

23. Madsen, H.T.; Bajraktari, N.; Hélix-Nielsen, C.; van der Bruggen, B.; Søgaard, E.G. Use of biomimetic forwardosmosis membrane for trace organics removal. J. Memb. Sci. 2015, 476, 469–474. [CrossRef]

24. Wijmans, J.G.; Baker, R.W. The solution-diffusion model: A review. J. Memb. Sci. 1995, 107, 1–21. [CrossRef]25. Madsen, H.T. Membrane Filtration in Water Treatment-Removal of Micropollutants; Elsevier: Amsterdam,

The Netherlands, 2014.26. Zhao, Y.; Qiu, C.; Li, X.; Vararattanavech, A.; Shen, W.; Torres, J.; Hélix-Nielsen, C.; Wang, R.; Hu, X.;

Fane, A.G.; et al. Synthesis of robust and high-performance aquaporin-based biomimetic membranes byinterfacial polymerization-membrane preparation and RO performance characterization. J. Memb. Sci. 2012,423–424, 422–428. [CrossRef]

Membranes 2016, 6, 46 12 of 12

27. Solvez, J.A.; Iniesta, A.; Garciáde la Torre, J. Radius of gyration of multisubunit macromolecules: Applicationto myosin heads, myosin rod and whole myosin. Int. J. Biol. Macromol. 1988, 10, 39–43. [CrossRef]

28. Bianchi, A.; Giorgi, C.; Ruzza, P.; Toniolo, C.; Milner-White, E.J. A synthetic hexapeptide designed to resemblea proteinaceous p-loop nest is shown to bind inorganic phosphate. Proteins Struct. Funct. Bioinform. 2012, 80,1418–1424. [CrossRef] [PubMed]

29. Case, D.A.; Babin, V.; Berryman, J.T.; Betz, R.M.; Cai, Q.; Cerutti, D.S.; Cheatham, T.E., III; Darden, T.A.;Duke, R.E.; Gohlke, H.; et al. AMBER 14; University of California: San Francisco, CA, USA, 2014.

30. Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.; Simmerling, C. Comparison of multiple amberforce fields and development of improved protein backbone parameters. Proteins Struct. Funct. Genet. 2006,65, 712–725. [CrossRef] [PubMed]

31. Jorgensen, W.L.; Chandrasekhar, J.; Madura, J.D.; Impey, R.W.; Klein, M.L. Comparison of simple potentialfunctions for simulating liquid water. J. Chem. Phys. 1983, 79, 926–935. [CrossRef]

32. Darden, T.; York, D.; Pedersen, L. Particle mesh Ewald: An N log(N) method for Ewald sums in large systems.J. Chem. Phys. 1993, 98, 10089–10092. [CrossRef]

33. Ryckaert, J.P.; Ciccotti, G.; Berendsen, H.J.C. Numerical integration of the Cartesian Equations of Motion of aSystem with Constraints: Molecular Dynamics of n-Alkanes. J. Comput. Phys. 1977, 23, 327–341. [CrossRef]

34. Roe, D.R.; Cheatham, T.E. PTRAJ and CPPTRAJ: Software for processing and analysis of molecular dynamicstrajectory data. J. Chem. Theory Comput. 2013, 9, 3084–3095. [CrossRef] [PubMed]

35. Gruber, M.F.; Wood, E.; Truelsen, S.; Østergaard, T.; Hélix-Nielsen, C. Computational Design of BiomimeticPhosphate Scavengers. Environ. Sci. Technol. 2015, 49, 9469–9478. [CrossRef] [PubMed]

36. Parisio, G.; Stocchero, M.; Ferrarini, A. Passive Membrane Permeability: Beyond the Standard Solubility-DiffusionModel. J. Chem. Theory Comput. 2013, 9, 5236–5246. [CrossRef] [PubMed]

37. Shinoda, W. Permeability across lipid membranes. Biochim. Biophys. Acta Biomembr. 2016, 1858, 2254–2265.[CrossRef] [PubMed]

38. Van der Bruggen, B.; Schaep, J.; Maes, W.; Wilms, D.; Vandecasteele, C. Nanofiltration as a treatment methodfor the removal of pesticides from ground waters. Desalination 1998, 117, 139–147. [CrossRef]

39. Nelson, D.; Cox, M. Lehninger Principles of Biochemistry, 4th ed.; W.H. Freeman and Company: New York, NY,USA, 2005.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open accessarticle distributed under the terms and conditions of the Creative Commons Attribution(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).